Artificial Intelligence Algorithms for Cost Estimation in Design Services: A Systematic Review

This study presents a systematic review and evaluation of Artificial Intelligence (AI) Algorithms applied to the estimation of engineering design service costs-a critical yet often underestimated component of construction project budgeting. Through a comprehensive analysis of recent literature, the research identifies and critically assesses AI-based approaches used in prior studies for forecasting costs associated with design, supervision, and preconstruction engineering activities. A three-criteria evaluation framework is employed to guide algorithm selection: (۱) predictive performance, measured by quantitative metrics such as Mean Absolute Percentage Error (MAPE), coefficient of determination (R۲), and accuracy; (۲) practical strengths and limitations, including robustness to data scarcity or noise, interpretability, computational efficiency, and susceptibility to overfitting; and (۳) prevalence and adaptability across diverse project types and contexts (e.g., building, infrastructure, and rehabilitation projects). Based on this systematic assessment, the study narrows the field to three high-performing and complementary algorithms Artificial Neural Networks (ANN), Random Forest (RF), and Extreme Gradient Boosting (XGBoost)-which collectively offer a balanced combination of accuracy, flexibility, and suitability for early-phase cost estimation under real-world constraints. The findings provide a methodological foundation for future AI-driven cost modeling in engineering services and support informed algorithm selection in both research and practice.

Keywords:

Cost Estimation , Engineering Service , Artificial Intelligence , Artificial Neural Network (ANN) , Extreme Gradient Boosting (XGBoost) , Random Forest (RF)